Orienting the Reader: What “Steps” Means in a CBGM Workflow

When students ask how the Coherence-Based Genealogical Method actually works, they usually expect a neat sequence from input to output. CBGM is indeed stepwise, but the steps form a loop rather than a straight line. Editors begin by collating witnesses and defining variation units. They then judge the direction of change within each unit by drawing a local stemma, compute “coherence” in two layers, identify potential ancestors, and watch global patterns emerge. Those global patterns are then used diagnostically to revisit local calls, refine the segmentation or the local arrows, recompute, and only then export a defended “initial text” for publication in the Editio Critica Maior (ECM) and, downstream, in handbook editions such as NA28 and beyond. This chapter walks steadily through that loop so a serious churchgoing Bible student can picture what the method counts, what it infers, and where documentary controls must be applied.

Step One: Gathering and Normalizing the Witness Evidence

CBGM stands on exhaustive collation. Editors assemble transcriptions of Greek manuscripts, add versional evidence where it is philologically secure, and register patristic citations that are precise enough to identify a reading. Normalization is required because scribes vary in orthography and spacing. Without a consistent baseline, a computer cannot count agreement responsibly. The normalization used in CBGM preserves meaningful differences while smoothing those that do not signal distinct readings. The aim is not to erase scribal individuality but to ensure that when the tool reports agreement, it is measuring the same thing across witnesses. This is the raw data that feeds everything else.

Step Two: Defining Variation Units With Care and Restraint

A variation unit is the smallest stretch of text where competing readings are in genuine competition. Unit boundaries must be drawn conservatively. If a single scribal act produced both a substitution and a word-order shift, treating them as separate units can create an illusion of independence that never existed in the copying event. Conversely, lumping unrelated differences into one unit blurs causal lines and muddies coherence. Segmenting the text is therefore a substantive judgment, not merely clerical work. CBGM’s later computations depend on these boundaries. This is the first place where a documentary mindset pays dividends, because editors who keep early witnesses in view often segment with an eye to the actual phenomena on the page rather than abstract theory.

Step Three: Enumerating Readings and Subreadings Within Each Unit

Once a unit is set, editors list the distinct readings. If two forms differ only in spelling without altering the lexical choice, they may be treated as subreadings. If the difference shifts sense or structure, it becomes a major reading. Attestation is then assigned: each witness is recorded as attesting one reading or another. The output of this step is descriptive. It claims nothing about originality. It simply states how the witnesses speak in this location. Beginners should learn to love this discipline. Description precedes evaluation. The accuracy of later inferences depends on whether the descriptive layer has been executed with precision.

Step Four: Measuring Pre-Genealogical Coherence Across the Book

Before any directional judgments are drawn, CBGM tallies agreement between witnesses across all defined units. This is pre-genealogical coherence. Two witnesses with high agreement display strong pre-genealogical coherence; two witnesses that often diverge display weak pre-genealogical coherence. Because this layer counts rather than infers, it has a welcome objectivity. No matter what an editor prefers, the numerical agreement stands. Strong pre-genealogical coherence often points toward clusters that deserve attention, such as the alignment frequently observed between late second–early third-century papyri and Codex Vaticanus (B, 300–330 C.E.) in certain books. Weak coherence can signal heavy contamination or simply gaps in a witness due to damage or lacunae. Either way, the student should grasp that, at this stage, nothing has yet been decided about which reading is earlier or later. We are only mapping how often witnesses say the same thing.

Step Five: Drawing Local Stemmata—The Editorial Hinge of the Method

The heart of CBGM is the local stemma. For each variation unit, editors judge which reading is prior and which readings are posterior, drawing arrows from earlier to later. This is where internal criteria legitimately operate, but always under the discipline of external evidence. A reading that obviously explains the rise of others, is attested by early and weighty witnesses, and coheres with the author’s language and the immediate context will be placed at the head. Posterior readings are linked to the prior by arrows that reflect plausible scribal tendencies such as harmonization, simplification, expansion, or assimilation to parallel passages. Because these micro-judgments are repeated across thousands of units, the cumulative effect shapes the method’s global portrait. A churchgoing student does not need to fear the technicalities here. The question is simple and concrete: in this unit, which reading could have produced the others, and who attests it earliest?

Step Six: Computing Genealogical Coherence From Local Arrows

Once local stemmata exist for a sufficient mass of units, CBGM aggregates their arrows. Genealogical coherence asks, across the entire set, whether a given witness tends to stand earlier relative to another. If Witness A holds the prior reading in a large proportion of units where Witness B holds a posterior reading, A emerges as a plausible source for B at the level of readings. The software represents this relation with a line from A to B. The thickness of the line reflects how many units support that direction; the thin line warns of fragility, the thick line signals persistent direction. Genealogical coherence does not claim direct copying, nor does it deny contamination. It simply says that, in the places where the two witnesses overlap, the pattern makes better sense if A’s readings are treated as antecedent to B’s.

Step Seven: Identifying “Potential Ancestors” and Building Substemmata

From genealogical coherence, the tool identifies “potential ancestors.” A potential ancestor of B is a witness that frequently supplies readings judged prior to B’s in overlapping units. For each witness, CBGM lists its closest potential ancestors and thereby builds a substemma, the set of witnesses that most plausibly stand earlier relative to it. This gives editors a reading-level lineage without asserting a physical exemplar chain. Because contamination is pervasive in New Testament copying, a single witness may have different potential ancestors in different sections. The substemma thus captures a realistic, mixed heritage rather than a rigid family tree. It is a flexible way to talk about influence without pretending to prove who copied whom.

Step Eight: Diagnosing Global Patterns and Stress-Testing Local Decisions

The global network that emerges after potential ancestors are computed functions as a diagnostic. If a witness that usually aligns with early anchors suddenly appears to feed a reading cluster that departs from those anchors, editors have reason to reexamine the relevant local stemmata. Sometimes the local arrows were drawn on assumptions that felt plausible in isolation but falter when set against the broader documentary picture. At other times the segmentation of the unit hid the causal link between related changes. The CBGM loop invites editors to return to the local evidence, reconsider the direction of change, adjust a boundary or an arrow if warranted, and then recompute. Responsible use of the method is iterative. It accepts feedback from the global portrait but never allows the portrait to override stubborn facts in early, high-quality witnesses.

Step Nine: Handling Gaps, Mixture, and Uneven Coverage With Transparency

Because witnesses are damaged, incomplete, or unevenly legible, CBGM must account for missing data. A witness may have no text for a unit where a critical decision lies. The tool neither penalizes nor rewards silence. It simply reduces overlap and, therefore, the power of any claimed direction between two witnesses. Mixture complicates matters further. A scribe may copy mostly from one exemplar while importing scattered readings from another. In a CBGM visualization, this will often appear as multiple potential ancestors with modest line weights rather than one dominant source. Beginners should read such graphs with patience. The unevenness is not a flaw in the method; it reflects the historical reality of how Christian copying actually occurred.

Step Ten: Incorporating Versions and Fathers Without Letting Them Overrule Greek Anchors

Versions and patristic citations expand the evidentiary footprint, especially where Greek witnesses are thin. CBGM can incorporate such evidence into both the descriptive and directional layers, provided the translation or citation is precise enough to point to one Greek reading rather than another. Yet translation habits and citation practices can mislead. A Latin rendering may collapse two Greek options; a father may paraphrase from memory. Editors therefore include versions and fathers with disciplined caution. Where the Greek manuscript base is strong—especially where early Alexandrian papyri and B agree—the versions and fathers are used to confirm spread, not to drive direction. Where the Greek base is thin, versions and fathers may carry more weight, but their contribution is explicitly bracketed by what we know about their reliability in the book and author under study.

Step Eleven: Exporting “Initial Text” Decisions to the ECM and Downstream Editions

After the iterative loop stabilizes, editors export, unit by unit, the reading judged to stand at the head of the tradition. In CBGM parlance, this is the “initial text.” Across a book, those decisions constitute the reconstructed text that the ECM prints and that handbook editions may adopt in whole or in part. The apparatus then records rival readings and their attestation. For the reader, this is where CBGM becomes visible as concrete changes or confirmations in passages they study. The path from raw collation to printed line is not short, and that is precisely why understanding the steps matters. Each printed decision rests on a definable chain of evidence, not on algorithmic mystique.

A Walk-Through With a Toy Unit: Three Readings, One Direction

Imagine a unit where three readings compete. Reading A is supported by an early papyrus and B (300–330 C.E.), reading B by a cluster of later witnesses, and reading C by a scattered mix. The editor asks which reading most plausibly gave rise to the others. Reading A, though shorter, makes excellent contextual sense and easily explains Reading B’s expansion. Reading C introduces a synonym that appears elsewhere in the author’s style but creates an awkward construction here. The local stemma places A at the head, with arrows to B and C. Across the book, the witnesses that favor Reading A in similar situations regularly stand earlier relative to those that favor B. Genealogical coherence therefore marks those A-leaning witnesses as potential ancestors of the B-leaning cluster. The ECM assigns A as the initial text, and NA adopts it. The reader sees a footnote, not a black box. The decision has been tracked from unit definition to global coherence and back.

How the Same Steps Play Out in Lived Cases: Jude 5

Jude 5 forces students to watch every step with special care. The major readings are “Jesus” and “Lord.” After defining the unit so that related differences are not artificially split, editors list attestation for each reading and draw a local stemma. The internal question is straightforward: which reading better explains the rise of the other? The documentary question is equally straightforward: which reading is supported by the earliest and best Greek witnesses? Genealogical coherence then aggregates similar decisions across Jude and its closest relatives, identifying potential ancestors for witnesses that carry each reading. Versions and patristic citations are consulted with the translation and citation caveats firmly in mind. The ECM’s initial-text assignment reflects that combined reasoning. A churchgoing reader who understands the steps can now interrogate the decision responsibly rather than being daunted by the label “coherence.”

How the Same Steps Play Out in Lived Cases: James 2:3

In James 2:3, the dispute involves word order that subtly colors the verse’s rhetoric. After careful segmentation, editors enumerate the readings and note that an early Alexandrian alignment, including B, supports one arrangement, while later witnesses prefer another. The local stemma is drawn by asking which arrangement a scribe would more likely create if attempting to smooth the sentence under social pressure toward rhetorical elegance. The prior reading will be the one that generates the others by plausible habits and is anchored by early witnesses. Genealogical coherence then reveals that the witnesses carrying the prior arrangement often function as potential ancestors for the later arrangement across adjacent units. The initial text is assigned accordingly. The end result for the Bible student is a translation nuance that rests on definable steps, not on the aura of an algorithm.

How the Same Steps Play Out in Lived Cases: 1 Peter 5:1

First Peter 5:1 revolves around the presence or absence of a connective. The unit is defined so that the particle’s status is assessed in context. Attestation reveals that the connective’s omission is widespread in later witnesses, while the presence appears in early, reliable attestations. The local stemma favors presence as prior because omission is a common scribal tendency when particles are perceived as superfluous. Genealogical coherence confirms that witnesses preserving the particle tend to stand earlier relative to those that omit it in overlapping units. The initial text is therefore assigned with the connective present. The decision shapes discourse flow in translation but does not strain doctrine, and its defense can be narrated in a few sentences for a Sunday school class because the steps are clear.

How the Same Steps Play Out in Lived Cases: 2 Peter 3:10

Second Peter 3:10 is the cautionary classroom example. The reading “not be found” lacks direct Greek manuscript support, yet the ECM adopted it as the initial text based on internal direction. The unit was segmented to capture the rival constructions; the local stemma was drawn to make sense of how the other readings could have arisen; genealogical coherence was computed; and the global portrait did not dissuade the local call. A documentary method treats such a result with explicit caution, since early Greek anchors do not support it. The steps are the same as in other units, but the distribution of evidence is different. Knowing the steps empowers the reader to say exactly why this decision sits in a different evidentiary posture from cases like James 2:3 or 1 Peter 5:1 where early Greek support is strong.

Internal Criteria Inside the Steps: Servants That Need Supervision

At Step Five, internal criteria such as intrinsic probability and transcriptional likelihood enter. CBGM gives these criteria a disciplined place to operate by forcing them to be expressed as arrows within concrete units rather than as free-floating preferences. Intrinsic arguments ask which reading best fits the author’s known usage; transcriptional arguments ask which reading a scribe would likely produce from another. Both are useful servants. Neither is a master. When early Alexandrian papyri such as P75 (175–225 C.E.) align with B in Luke and John, those anchors cannot be overridden by internal attractions without compelling reasons stated in the local stemma. Where early anchors are absent or divided, internal criteria may carry more explanatory weight, but their force remains explicitly proportioned to the documentary situation.

Thresholds, Weights, and Why CBGM Is Not a Black Box

Students sometimes fear that unseen thresholds determine everything. In practice, the thresholds used to flag potential ancestors or to display line-weight reflect editorially transparent choices about how much overlapping evidence is needed to claim a directional relation. If the overlap is small, any claim is fragile and is displayed accordingly. If the overlap is broad and the local arrows consistently favor one direction, the relation grows more persuasive. None of this is hidden magic. It is simply the mathematics of tallying many small judgments and letting the weight of numbers reflect the strength of the case. The wise reader lets line thickness guide attention without letting it silence questions about the underlying evidence in key units.

What the Steps Do Not Do: They Do Not Erase Early Documentary Anchors

Following the steps can mislead the unwary into thinking that once coherence is computed, early witnesses no longer matter. The opposite is true. CBGM’s birth purpose was to describe relationships honestly in a contaminated tradition. Its steps are a discipline for organizing and testing judgments, not a license to ignore the earliest, most reliable testimony. When P66 or P75 or B stands in clear agreement for a reading in John, or when an early papyrus and B align in Luke, those anchors deserve first attention. The steps then help explain why that alignment makes historical sense and how later readings likely arose. Where those anchors are absent, the same steps enable a careful, provisional reconstruction that remains open to fresh discovery.

Iteration as a Virtue, Not a Vice: Why Recomputing Is Part of the Method

Because local stemmata feed the global portrait and the global portrait feeds back into local re-evaluation, CBGM is intentionally iterative. Recomputing after a refined segmentation or a corrected local arrow is not an admission of failure; it is how the method protects itself from premature certainty. The loop moves toward stability as more units are decided with clarity, as early anchors are weighted responsibly, and as versional and patristic evidence is calibrated for reliability. The output is not a single leap to a finished text but a set of defended, revisable decisions that can be tested whenever new data or better arguments appear.

How the Steps Interact With Editorial Prudence in NA-Era Decisions

The ECM’s decisions feed into handbook editions cautiously. Not every CBGM-supported initial text is adopted wholesale. Editors of a handbook edition consider readability, user familiarity, and the broader consensus of the manuscript tradition alongside the ECM. Where CBGM confirms a reading already supported by early anchors, adoption is straightforward. Where CBGM elevates a reading on internal grounds against the grain of Greek evidence, editors may hesitate, annotate more heavily, or choose to wait for further study. Understanding the steps helps the reader see why editions occasionally differ and why apparatus notes grow more informative in places where judgments are delicate.

Reading CBGM Outputs Without Graphs: A Plain-Speech Strategy

Most churchgoers will never draw a CBGM graph, nor do they need to. The steps can be narrated in plain speech whenever a variant note appears. One asks what the rival readings are, which early witnesses attest each, which reading better explains the rise of the others in that unit, and whether the witnesses that carry the prior reading tend to stand earlier relative to others across adjacent units. If the answer to these questions points steadily in one direction, confidence grows that the printed reading reflects the earliest recoverable text. If the answers are divided, confidence is proportioned accordingly, and translation may footnote both options without alarm. This is the practical payoff of understanding the steps. Confidence is tethered to evidence rather than to slogans.

A Note on Byz Inside the Steps: A Computed Cluster, Not a Trump Card

When CBGM visualizations label a large cluster “Byz,” they signal high agreement among later witnesses in the database’s terms. Within the steps, this computed cluster functions as an empirical description of late uniformity. It neither guarantees originality nor proves lateness in any given unit. The local stemma still decides direction, and early external anchors still discipline the judgment. The presence of a large, coherent late cluster can, however, alert editors to the probability that a reading’s spread is due to liturgical smoothing, harmonization, or standardization. Within the steps, that alert prompts a careful test, not a foregone conclusion.

Why These Steps Suit a Documentary-First Approach

A documentary approach insists on recovering the original wording by weighing the earliest, most reliable witnesses first and letting internal criteria serve that goal. CBGM’s steps can be used in precisely that spirit. Pre-genealogical coherence helps locate where early witnesses speak together; local stemmata let internal reasoning be stated transparently and tested; genealogical coherence and potential ancestors trace how early readings likely flowed into mixed later witnesses; and the diagnostic loop prevents local decisions from drifting away from the global manuscript reality. Used this way, CBGM does not compete with the documentary method. It becomes a tool that organizes its findings and exposes where internal arguments need either confirmation from, or correction by, the early evidence.

Bringing the Steps Together in One Unbroken Narrative

To see the whole process as a single flow, picture editors opening a book of the New Testament and collating every witness they can responsibly transcribe. They define each place where readings compete, enumerate those readings with their attestation, and count agreements to see who often speaks together. They then draw thousands of tiny arrows inside local stemmata, arguing in each case for which reading is prior based on documentary anchors and sober internal tendencies. They aggregate those arrows to see which witnesses routinely stand earlier relative to others, flag those as potential ancestors, and watch a global network take shape. That network is not the last word; it is a mirror that sends them back to recheck questionable units, refine segmentations, and redraw a few arrows. Only when the loop stabilizes do they export an initial text, print an apparatus that lets others test their work, and invite the church’s scholars and careful readers to probe, confirm, or challenge what has been done—always with the manuscripts in hand.

You May Also Enjoy

The Book of Acts and Textual Criticism: Evaluating the Manuscript Evidence and Restoring the Original Text

About the Author

CLICK LINKED IMAGE TO VISIT ONLINE STORE

CLICK TO SCROLL THROUGH OUR BOOKS